Early warning technology for cracking severity based on improved cultural differential algorithm

Effective monitoring and early warning for cracking severity are directly related to the ethylene production stability and the overall economic benefits. Elman neural network is used to establish the early warning model of cracking severity, and an improved Cultural Differential Evolution algorithm(ICDE) is proposed to training the model. This hybrid algorithm integrates Cultural algorithm(CA) and Differential Evolution algorithm(DE) together, and increases some measure indicators of evolution direction, to make the algorithm more adaptive. Simulation results show that this method gains accurate warning signals to the cracking working conditions timely and effectively, obtained satisfactory results.